BEHAVIOUR & INFORMATION TECHNOLOGY, 1991, VOL. 10, NO. 4, 257-266.
Reading and skimming from computer screens and books: the paperless office revisited?
PAUL MUTER and PAULA MAURUTTO
Psychology Department, University of Toronto, Toronto, Ontario, M5S 3G3, Canada
Address correspondence to Paul Muter, Department of Psychology, University of Toronto, Toronto, Ont., M5S 3G3, Canada firstname.lastname@example.org
(c) Copyright 1991 Taylor and Francis Ltd.
Abstract. Past research has demonstrated that reading efficiency is lower from standard computer displays of the 1980s than from paper. In the present experiments, subjects read or skimmed stories, sometimes from a high-quality CRT (cathode ray tube) and sometimes from a book. Skimming was 41% slower from the CRTs than from the book. Possible reasons for this finding are discussed. Reading speed and comprehension were equivalent for the high-quality CRTs and the book. The paperless office may be imminent after all.
In the 1970s, remarkable developments in technology resulted in the prediction of a paperless society in the not too distant future (Lancaster 1978). However, several experiments have indicated that reading is slower from CRTs than from paper (e.g., Muter et al. 1982). This may, in part at least, account for the fact that, contrary to expectations, the widespread increase in computer use has been associated with an increase in paper output and a change in predictions: "paper ... is still the most popular method of communications and is likely to remain so" (Plume 1988). Before a paperless society can emerge, reading must be at least approximately as efficient from CRTs as it is from paper. In the present article, recapitulation of these early studies will be conducted in order to ascertain whether advancements in displays coupled with variations in textual format can produce similar reading efficiency (speed and comprehension) for CRTs and books.
Muter et al. (1982) compared speed and comprehension in reading from a videotex terminal and a book. Results over two hours of reading indicated that, though extended reading from videotex was feasible, it was 28.5% slower than reading from paper. There was no difference in comprehension. Other researchers have also found decreased efficiency from CRTs (Creed et al. 1987; Gould and Grischkowsky 1984, 1986; Gould et al. 1987a; Haas and Hayes 1986; Hansen and Haas 1988; Kruk and Muter 1984; Wilkinson and Robinshaw 1987; Wright and Lickorish 1983; see Mills and Weldon 1986, for a review). Some of these experiments involved reading, and some proofreading. Reading and proofreading share some component processes, but other processes are unique to each skill.
Table 1 contains a partially overlapping list of usual differences between reading from a book and reading from a computer that could conceivably account for a difference in reading efficiency. Research has suggested effects of some of these variables; for example: characters per line (Duchnicky and Kolers 1983); polarity (Bauer and Cavonius 1983); resolution (Harpster et al. 1989); interline spacing (Kruk and Muter 1984); justification (Campbell et al. 1981; Trollip and Sales 1986); and chromaticity (Matthews et al. 1989). More research is needed to determine the main effects and the interactions of the variables in Table 1.
Some Usual Differences between Book and Computer Reading
With regard to proofreading speed, Gould et al. (1987b) reported no significant difference between CRTs and paper using a high resolution, anti-aliased display on which fonts resembled those on paper. Similar findings were reported for reading by Oborne and Holton (1988), though they achieved this result by creating a degraded paper condition that resembled the CRT condition, rather than creating an enhanced CRT condition.
In addition to reading, skimming was studied in Experiment 1. It has been estimated that approximately 30% of the activity of skilled readers can be characterized as skimming (Masson 1982). Furthermore, the role of skimming from CRTs has become increasingly important given the explosion of information and the widespread use of computerized information retrieval from electronic mail, computerized abstracts, online technical manuals, etc. Skimming may be simply the skill of reading sped up, or it may be a fundamentally distinct process. Very little research has been done on skimming, and we know of no previous work comparing skimming from CRTs to skimming from paper.
2. Experiment 1
In Experiment 1, a higher quality display than in the earlier reading experiments was used to test the hypothesis of no difference in speed or comprehension between CRTs and normal book conditions for either reading or skimming. The phrase "normal book conditions" is used to refer to the textual format which is typically found in books. Interline spacing, characters per line, etc, were not equivalent across the two conditions because the purpose of this experiment was to determine whether CRT displays can be improved so that reading speed and comprehension are equivalent to those from standard books. (Features such as double spacing entail little if any additional cost on a CRT, whereas they are costly in books.)
Skimming was defined as proceeding at a rate three to four times faster than normal reading "in order to grasp a general sense of the content or to retain only the main points."
Twenty-four subjects, 10 females and 14 males, responded to public announcements. Nineteen were students at the University of Toronto; the remaining five were full time employees. Ages ranged from 19 to 30. All subjects were required to be able to read English fluently, though five subjects had learned English as their second language. Thirteen participants wore corrective lenses.
2.1.2 Materials and Apparatus
Short stories from H.H. Munro, The Complete Works of Saki (1976), were used as the reading materials. The book condition consisted of approximately five pages per story on average, with text covering an area of 10.5 by 17 cm on the page. There were approximately 41 rows of text per page with 60 characters per row. The 10-point text was single spaced, right and left justified.
In the CRT condition, a Macintosh™ IIx computer was used with an AppleColor™ high-resolution RGB display. This monitor contains 640 (horizontal) by 480 (vertical) picture elements. The vertical refresh rate is 66.7 Hz. The particular monitor used in Experiment 1 had an active video area of 235 mm X 176 mm, so the resolution was approximately 69 dpi. The characters were 12-point. The brightness knob was set at the point suggested by the Macintosh manual as the optimal setting for viewing the screen. There was no anti-aliasing.
The CRT condition included approximately 12 pages per story, with text encompassing an area of 20.5 X 13.5 cm. On average there were approximately 12 rows of text per page.
Included in the CRT condition were various "enhancements" which might improve reading efficiency. These enhancements were included to help determine whether reading from a CRT can be as efficient as reading from a normal book. Enhancements included: double spacing (Kruk and Muter 1984); negative contrast, i.e., black characters on a white background (Bauer and Cavonius 1983); Chicago font, which is boldfaced (Krulee and Novy 1986); proportional spacing of characters within words (Beldie et al. 1983); three-space indentation of every other line to facilitate return sweeps (Huey 1908); left justification only (Trollip and Sales 1986); 85 characters per line; eight-space indentation of the first line of each paragraph; and horizontal separation of three spaces between sentences.
The room was illuminated from an overhead light source, in addition to a reading lamp which subjects could adjust in whatever manner they preferred.
2.1.3. Design and Procedure
Half of the subjects were randomly allocated to the reading task and half to the skimming task. Condition, book or CRT, was a within-subject variable, with three stories read or skimmed from the book and three from the CRT in a strictly alternating order. Half of the subjects read from the book on the first trial and half from the CRT. Story order was randomly determined for each subject.
Subjects were informed that they would be reading or skimming from both the CRT and the book and that their reading speed, comprehension, and distance from the screen would be monitored. They were told that they were not permitted to turn back to previous pages in either condition. Participants were allowed to hold the book in whatever manner they preferred and to adjust the distance from the screen.
In the skimming task, subjects were asked to proceed at a rate three to four times faster than normal "in order to grasp a general sense of the content or to retain only the main points." These instructions were repeated to subjects before each story. A pretest was given to those subjects who skimmed in order to give them an indication of the speed at which they were expected to skim. These subjects were told to read an instruction sheet while their time was monitored. They were then asked to skim the same sheet in one quarter of the time originally taken to read it. (A pilot study had indicated that, without these instructions, subjects who were asked to skim would proceed at approximately their normal reading rate.)
In the book condition, reading and skimming speeds were monitored by the experimenter through the use of a stop watch. Subjects informed the experimenter when they were finished reading. In the CRT condition, speed was calculated by the computer program. To get to the next screen page, subjects pressed a mouse; there was no scrolling. The screen was cleared and a new screenfull was presented each time the mouse was pressed. Approximately one minute after a trial began, the experimenter estimated the distance from the reading material to the subjects' eyes.
Following each story a comprehension test was administered consisting of 10 short answer questions presented in random order for 10 seconds each. Subjects wrote down the question number and the answer. At the end of the experiment, subjects were asked to rate their preference for either the book or the CRT on a seven-point scale.
A within-subject 2 X 3 (medium X trial) analysis of variance on reading speed indicated no significant difference between reading on a CRT versus reading from paper, F(1, 11) = 2.04; see Table 2. Of the 12 subjects, four read faster on the CRT and eight read faster from the book. Comprehension was higher in the CRT condition, but this difference did not reach significance, F(1,11) < 1. Of the 12 subjects, eight had greater comprehension in the CRT condition, and four had greater comprehension in the book condition. The effective reading rate (reading rate times proportion correct on the comprehension test; Jackson and McClelland 1979) was actually slightly, but not significantly, higher in the CRT condition than in the book condition, F(1, 11) < 1. An analysis of statistical power (1 - beta) for effective reading rate revealed that the probability is greater than 0.9 that Experiment 1 would have detected a significant effect (alpha = .05) if the true mean treatment effects were 10%. Thus, Experiment 1 was quite sensitive.
Skimming speeds, on the other hand, proved to be significantly different between the book and CRT conditions, F(1, 11) = 7.44, p < .05. All 12 subjects skimmed slower from the CRT than from the book. Though subjects skimmed 41% slower from the CRT, their comprehension was better, F(1, 11) = 7.23, p < .05. Nine of the 12 subjects answered more questions correctly from the CRT than from the book. Apparently, there was a speed-accuracy tradeoff. The effective skimming rate (skimming rate times proportion correct on the comprehension test) was 140.8 in the CRT condition and 179.6 in the book condition, F(1, 11) < 1.
Comparing subjects who skimmed with those who read, the overall average skimming rate (M = 676 words per min) was 3.3 times faster than the average for the subjects in the reading task (M = 205 words per min). However, fewer questions were answered correctly in the skimming condition (M = 2.46) than in the reading condition (M = 4.97).
The mean preference rating for reading was 4.3, where 7 indicated a strong preference for the CRT and 1 indicated a strong preference for the book. The mean preference rating for skimming was 4.6. Fourteen of the 24 subjects, seven for each task, preferred reading or skimming from the CRT.
For reading, distance from the reading material was greater in the CRT condition (M = 58.6 cm) than in the book condition (M = 48.0 cm), F (1, 11) = 7.63, p < .05. For skimming, there were no significant effects on distance.
3. Experiment 2
In Experiment 1, the skimming results were quite clear. However, the reading results tended to support the null hypothesis. Experiment 2 was designed to replicate the reading results under slightly different conditions with eighteen new subjects. A second purpose of Experiment 2 was to compare "enhanced" formatting of text in the CRT condition to more standard formatting. The skimming conditions were not included in Experiment 2.
Eighteen subjects between the ages of 19 and 29 were tested, 12 males and 6 females. Subjects either responded to announcements and were paid, or were psychology students who volunteered for credit. Nine subjects wore corrective lenses, and for one subject English was a second language.
3.1.2. Materials and Apparatus
The six Saki short stories were similar to those in Experiment 1 except that they were shorter.
In Experiment 2, a Radius Full Page Display™ was used in the CRT conditions. The active video area is approximately 210 mm (horizontal) by 285 mm (vertical). The Radius low-curvature screen has 640 horizontal X 864 vertical picture elements (approximately 77 dpi). The vertical refresh rate is 69 Hz. This monitor presents black characters on a white background.
In the CRT conditions, two types of textual display formats, called A and B, were used. The format for display A was designed to resemble the format typically found on many personal computer screens of the 1980s. It consisted of a Monaco font, 12-point, with nonproportional spacing of characters within words. Text was single spaced with a blank line between paragraphs. The first line of each paragraph was indented three spaces. The text covered a space of approximately 17 cm (height) X 18 cm (width). There were approximately 33 rows of text per page, with a maximum of 80 characters per row. There were on average approximately four pages per story.
Format B was similar to the format in the CRT conditions in Experiment 1. It included double spacing, and three lines separating paragraphs, with the first line of each paragraph indented eight spaces. Every other line was indented three spaces. Text consisted of a 14-point Chicago font, which is boldfaced with proportional spacing of characters within words. Text covered on average an area of 24 cm (height) X 18 cm (width), with approximately 21 rows of text per page and the same number of characters per row as in Display A. On average there were approximately six pages per story.
In both conditions the luminance contrast was set at the maximum. Experiment 2 was run in a different laboratory than Experiment 1.
3.1.3 Design and Procedure
Each subject read six stories, and received all three conditions twice. Each of the six possible orders of the three conditions was used equally often across the eighteen subjects. An example of the order of trials for a subject is: CRT-A, CRT-B, book, CRT-A, CRT-B, book. In other words, the order of trials for a subject's second replication was the same as for his or her first replication. The order of the stories was randomly determined for each subject.
Instructions, monitoring of speed and distance, procedures for advancing to the next screen, and the form of the comprehension tests were the same as in Experiment 1. Subjects were asked to indicate, on a seven-point scale, whether they preferred the book or the CRT, and whether, within the CRT condition, they preferred format A or B.
A within-subject 3 X 2 (condition X trial) analysis of variance on reading speed indicated no significant difference among the three conditions, F(2, 34) < 1; see Table 3. Seven of the 18 subjects read fastest in the book condition; nine subjects read fastest in condition CRT-A; and two read fastest in condition CRT-B. There were no significant differences in comprehension scores, F(2, 34) = 1.01, or effective reading rates, F(2, 34) = 1.26. An analysis of statistical power for effective reading rate revealed that the probability is greater than 0.9 that Experiment 2 would have detected a significant effect (alpha = .05) if the true mean treatment effects were 10%.
A within-subject 3 X 2 (condition X trial) analysis of variance on distance between the subject and the reading material indicated a significant effect of condition, F(2, 34) = 15.89, p < .01. The means were 29.8 cm, 43.0 cm, and 52.5 cm in the book, CRT-A, and CRT-B conditions respectively. (We have no explanation for the difference in distance between Experiments 1 and 2 with regard to the book condition. Perhaps it had something to do with the difference in laboratories, e.g., different chair and lighting.)
Nine of the 18 subjects preferred the book condition over the CRT conditions and seven subjects preferred the CRT conditions; two had no preference. The mean preference rating was 4.1, where seven indicated a strong preference for the CRT, and one indicated a strong preference for the book. Thirteen of the 18 subjects preferred display B over display A; the mean preference rating was 5.1, where seven indicated a strong preference for CRT-B.
These experiments demonstrate that reading from computer screens that are readily available in 1991 can be equivalent in speed and comprehension to reading from a book. The paperless office may be imminent after all. The increase in reading speed in comparison to earlier research may be attributable to the quality of the screen and the clarity of the characters. The advancements in computer technology, for example in resolution, and clearer and more varied fonts, have increased the legibility of the screen and allowed for more flexibility in textual presentation of information. Negative contrast capability (dark characters on a light background) increases the risk of the perception of flicker, but the flicker problem can be solved with a high refresh rate. Negative contrast reduces optical distortions, and increases acuity, contrast sensitivity, speed of accommodation, and depth of field (Bauer and Cavonius 1983; but see Taylor and Rupp 1987). It also decreases the problem of interfering reflections of external light (Bauer 1987).
The lack of a significant difference in reading efficiency was obtained in spite of the fact that the statistical power of the present experiments was quite high, especially in Experiment 2, which involved 18 subjects reading six stories each. In both experiments, medium (book or CRT) was a within-subject variable. Both experiments were sufficiently sensitive to yield significant effects of distance, and Experiment 1 was powerful enough to reveal a significant difference between reading and skimming, even though this was a between-subjects variable.
Skimming speed was markedly different for the two media. Skimming in the book condition was 41% faster than skimming in the CRT condition in Experiment 1. This result is probably not a serious impediment to the paperless office for at least four reasons. First, the difference may be attributable to the format of the textual display rather than to any screen deficit. The textual format in the CRT condition may have been closer to optimal for reading than for skimming, which may be a fundamentally different process. The selection of the CRT format was based on studies examining reading at normal rates, not skimming. Kolers et al. (1981) suggested that narrow columns might improve skimming because they eliminate the need for lateral eye movements. The shorter line lengths in the book condition, 60 characters compared to 85 characters in the screen condition, may have facilitated skimming from the book.
Second, the skimming speed difference may have been largely the result of a speed-accuracy tradeoff. For skimming, comprehension was higher from the CRT than from the book. The difference in the effective skimming rate was much smaller than the difference in the uncorrected skimming rate.
Third, the difference may have been related to the greater number of words per page in the book condition (Hansen and Haas 1988). For example, perhaps subjects have some tendency to consume pages at a fixed rate.
Fourth, perhaps the results were specific to the task. In this experiment, subjects were asked to obtain a general sense of what the story was about. If the task were altered to skimming for specific information, perhaps different results would be obtained.
In most previous experiments in this area, distance between the reader and the reading material was neither controlled nor reported. Evidence suggests that, within reasonable limits, distance between the reader and the reading material has no effect on perceptual span (Morrison and Rayner 1981) or reading efficiency (Kruk and Muter 1984). This is probably because, with increasing distance, though retinal image size decreases linearly, so does retinal eccentricity (distance of the image from the fovea), and these two effects offset each other exactly. (Acuity is a linear function of eccentricity; Anstis 1974). Similarly, within reasonable limits, size of the characters appears to have no effect on proofreading speed (Gould and Grischkowsky 1986), probably for the same reason. Nonetheless, the present data suggest that readers have a preference for a certain optimal visual angle of the characters, and they tend to spontaneously adjust the reading distance accordingly: Across conditions, distance tended to increase with the size of the characters.
In Experiment 2, textual format had no effect. It is possible that one or two of our "enhancements" had a negative effect which neutralized any positive effect of the remaining enhancements, or that two or more of the variables interacted in a negative way. The greater number of words per page may have favored CRT-A (Hansen and Haas 1988). Further experiments may reveal a format that will improve reading efficiency from computer screens. Computers offer the additional opportunity of increasing reading efficiency by means of dynamic text presentation (Juola 1988; Kang and Muter 1989; Muter et al. 1988; Williamson et al. 1986).
In our opinion, a high-quality screen permitting efficient reading is a necessary but not sufficient condition for a paperless office. Also probably necessary are: several megabytes of RAM (random access memory); multiple windows and a large screen (to simulate a large desktop; see Lansdale 1988); several dozen megabytes of secondary storage; an excellent user interface; and high-speed communications capability. All of these are now available within the price range of many computer users.
This work was supported by Natural Sciences and Engineering Research Council of Canada grants U0149 and G1779 to the first author. We thank Marika Jukutas and Linda Tilley for assistance, and Patrick J. Bennett, Mary Anne Buttigieg, Robert L. Duchnicky, Richard S. Kruk, Paul W. Smith, and Paul H.N. Yee for useful comments. Experiment 1 was conducted in the laboratory of Patrick J. Bennett, University of Toronto.
Anstis, S.M. 1974, A chart demonstrating variations in acuity with retinal position, Vision Research, 14, 589-592.
Bauer, D. 1987, Improving the VDU workplace by introducing a physiologically optimized bright-background screen with dark characters: advantages and requirements. In B. Knave and P.G. Wideback (Eds.), Work with display units 86 (Amsterdam: North Holland).
Bauer, D. and Cavonius, C. R. 1983, Improving the legibility of visual display units through contrast reversal. In E. Grandjean and E. Vigliani (Eds.), Ergonomic Aspects of Visual Display Terminals (London: Taylor and Francis).
Beldie, I.P., Pastoor, S., and Schwarz, E. 1983, Fixed versus variable letter width for televised text, Human Factors, 25, 273-277.
Campbell, A.J., Marchetti, F.M., and Mewhort, D.J.K. 1981, Reading speed and text production: A note on right-justification techniques, Ergonomics, 24, 633-640.
Creed, A., Dennis, I., and Newstead, S. 1987, Proof-reading on VDUs, Behaviour & Information Technology, 6, 3-13.
Daniel, D.B. and Reinkin, D. 1987, The construct of legibility in electronic reading environments, In D. Reinking (Ed.), Reading and Computers: Issues for Theory and Practice (New York: Teachers College Press).
Duchnicky, R. L., and Kolers, P. A. 1983, Readability of text scrolled on visual display terminals as a function of window size, Human Factors, 25, 683-692.
Gould, J. D., Alfaro, L., Barnes, V., Finn, R., Grischkowsky, N., and Minuto, A. 1987a, Reading is slower from CRT displays than from paper: attempts to isolate a single-variable explanation, Human Factors, 29, 269-299.
Gould, J. D., Alfaro, L., Finn, R., Haupt, B., and Minuto, A. 1987b, Reading from CRT displays can be as fast as reading from paper, Human Factors, 29, 497-517.
Gould, J. D. and Grischkowsky, N. 1984, Doing the same work with hard copy and with cathode-ray tube (CRT) computer terminals, Human Factors, 26, 323-337.
Gould, J.D. and Grischkowsky, N. 1986, Does visual angle of a line of characters affect reading speed?, Human Factors, 28, 165-173.
Haas, C. and Hayes, J.R. 1986, What did I just say? Reading problems in writing with the machine, Research in the Teaching of English, 20, 22-35.
Hansen, W.J. and Haas, C. 1988, Reading and writing with computers: A framework for explaining differences in performance, Communications of the ACM, 31, 1080-1089.
Harpster, J.L., Freivalds, A., Shulman, G.L., and Leibowitz, H.W. 1989, Visual performance on CRT screens and hard-copy displays, Human Factors, 31, 247-257.
Huey, E. B. 1908/1968, The psychology and pedagogy of reading (Massachusetts: M.I.T. press).
Jackson, M.D. and McClelland, J.L. 1979, Processing determinants of reading speed, Journal of Verbal Learning and Verbal Behavior, 108, 151-181.
Juola, J.F. 1988, The use of computer displays to improve reading comprehension, Applied Cognitive Psychology, 2, 87-95.
Kang, T.J. and Muter, P. 1989, Reading dynamically displayed text, Behaviour & Information Technology, 8, 33-42.
Kolers, P. A., Duchnicky, R. L., and Ferguson, D. C. 1981, Eye movement measurement of readability on CRT displays, Human Factors, 23, 517-527.
Kruk, R. S., and Muter, P. 1984, Reading of continuous text on video screens, Human Factors, 26, 339-345.
Krulee, G. K., and Novy, F. 1986, Word processing and effects of variability in type font, Perceptual and Motor Skills, 62, 999-1010.
Lancaster, F.W. 1978, Towards paperless information systems (New York: Academic Press).
Lansdale, M.L. 1988, The psychology of personal information management, Applied Ergonomics, 19, 55-66.
Masson, M. E. J. 1982, Cognitive processes in skimming stories, Journal of Experimental Psychology, 8, 400-417.
Matthews, M.L., Lovasik, J.V., and Mertins, K. 1989, Visual performance and subjective discomfort in prolonged viewing of chromatic displays, Human Factors, 31, 259-271.
Mills, C.B. and Weldon, L.J. 1986, Reading text from computer screens, ACM Computing Surveys, 19, 329-358.
Morrison, R. E. and Rayner, K. 1981, Saccade size in reading depends upon character spaces and not visual angle, Perception & Psychophysics, 30, 395-396.
Munro, H. H. 1976, The complete works of Saki (New York: Doubleday and Company).
Muter, P., Kruk, R., Buttigieg, M.A., and Kang, T.J. 1988, Reader-controlled computerized presentation of text, Human Factors, 30, 473-486.
Muter, P., Latremouille, S. A., Treurniet, W. C., and Beam, P. 1982, Extended reading of continuous text on television screens, Human Factors, 24, 501-508.
Oborne, D. J., and Holton, D. 1988, Reading from screen versus paper: There is no difference, International Journal of Man-Machine Studies, 28, 1-9.
Plume, T. 1988, Practical applications of optical disk image systems in document management, In P.A. Yates-Mercer (Ed.), Future Trends in Information Science and Technology (London: Taylor Graham).
Stewart, T.F.M. 1979, Eyestrain and visual display units: a review, Displays, 25-32.
Taylor, S.E. and Rupp, B.A. 1987, Display image characteristics and visual response. In B. Knave and P.G. Wideback (Eds.), Work with display units 86 (Amsterdam: North Holland).
Trollip, S.R. and Sales, G.S. 1986, Readability of computer-generated fill-justified text, Human Factors, 28, 159-163.
Wilkins, A. 1986, Intermittent illumination from visual display units and fluorescent lighting affects movements of the eyes across text, Human Factors, 28, 75-81.
Wilkinson, R.T. and Robinshaw, H.M. 1987, Proof-reading: VDU and paper text compared for speed, accuracy and fatigue, Behaviour & Information Technology, 6, 125-133.
Williamson, N.L., Muter, P. and Kruk, R.S. 1986, Computerized presentation of text for the visually handicapped, In E. Hjelmquist and L.G. Nilsson (Eds.), Communication and Handicap: Psychological Compensation and Technical Aids (North-Holland: Elsevier).
Wright,P. and Lickorish, A. 1983, Proof-reading texts on screen and paper, Behaviour & Information Technology, 2, 227-235.
Wright,P. and Lickorish, A. 1984, Investigating referees' requirements in an electronic medium, Visible Language, XVIII, 186-205.